RU2649304C2 - Method and system for effective compiling of media content elements for media platform upon request - Google Patents

Abstract

FIELD: data processing.

SUBSTANCE: invention refers to the method and system for compiling media content elements for the media platform upon request for rental by users of the media platform upon request. Proposed method uses the combination of the advisory system in order to select the appropriate set of media content elements to be offered during the next service period, and the predictor of the number of rents in order to assess how many videos the individual users will rent in the next service period. In addition, the method and system can be performed in order to estimate the profit or loss from leasing for the specified service period, as well as to assess customer satisfaction.

EFFECT: technical result is the automatic compilation of media content elements for the media platform upon request.

15 cl, 1 dwg

Description

BACKGROUND

The invention relates to a method and system for efficiently compiling media content elements for an on-demand media platform for users to rent an on-demand media platform and takes into account the following environment. An on-demand media operator (for example, a cable television or communications operator) offers on-demand media content elements to a group of end users, where end users can request rental of media content elements from a given set of currently available media content elements at any convenient time , for example, through a website, over the phone or a suitable television channel.

Elements of media content can be physical objects, such as DVDs, sent to end users by mail, or digital, as audio-video files broadcast from media servers upon request to end users via a data network. Users can either pay for the unit of the media content item that they rented (rent), or they can have a subscription to rent as many media content items as they want from the many currently available media content items (unlimited tariff) . In alternative payment models, the end user can pay a monthly fee, such as a monthly payment, and an additional fee for a rented media content item (subscription cost). In this alternative payment model, a monthly payment may cover a predetermined amount or cost of media content elements that the end user can rent without charging an additional fee (limited unlimited tariff).

Typically, the set of currently proposed media content items, called the "item set" for short, is fixed for a specific period of time, hereinafter referred to as the "maintenance period". The service period may be, for example, a week or a month. The operator can also offer their services as an Internet service, not limited to this cable, telecommunications or logistics network.

In order to effectively satisfy users of the media platform on demand, the operator must re-determine what elements of media content he wants to offer users for the next service period. An operator must rent certain elements of media content from one or more content providers. The content provider (p) typically offers various sets of elements (S (p)) of media content elements for the operator, called S1 (p), S2 (p), ..., Sn (p). The operator selects from each content provider zero or more sets of elements to offer end users for the next period. To be able to offer elements of media content, the operator is in negotiations with one or more content providers.

Various financial arrangements may exist between the content provider and the operator. For example, an operator may pay a fixed price for making this set of elements available, where the price may depend on the number of media content elements in the set of elements, their popularity, their quality and their year of release. In addition, the number of subscribers (that is, end users) that the operator serves can influence this fixed price. Alternatively, there may be a rental payment model where the operator must pay a certain amount of money to the content provider each time an element of media content is leased by the end user. This amount may be the same for all media content elements in the set of elements, but may also depend on the media content element (for example, it may be higher if it is a recently released feature film). Another possibility is that the financial agreement consists of paying a fixed price and an additional fee for each time an element of media content is rented by the end user.

Similarly, the financial agreement between the operator and the end user may concern a fixed subscription fee, regardless of the number of media content elements that the end user will rent in the next period, or it may be a type of rental payment, where the payment may depend on the corresponding element media content. The latter case, where the amount of payment depends on the element of media content, is called the differential rent model. Alternatively, the end user pays a monthly fee for a predetermined period of time, and additionally he must pay an additional fee for each element of media content that he or she rents.

When choosing which set of elements to choose, the operator should take into account the following questions. In order to satisfy end users of the media platform on demand, the goal is to offer a wide range of media content elements, in a variety to cover the different interests of each end user, and in quantity for each end user to find a sufficient number of media content elements that he or she wants to rent. If the number of videos submitted and their quality are not sufficiently appreciated by end users, end users are likely to rent less videos and consider unsubscribing, which will result in lower revenue. In the rental-end-user subscription model, the available media content elements should be of interest, so that sufficient income is generated to pay for the services of the content providers and a reasonable profit is created. In a fixed monthly fee model, end users should be satisfied with the proposed set of elements, so that they will renew their subscription, and so that other end users are likely to join.

On the other hand, in order to avoid losses, the operator must balance the breadth of the set of elements of the proposed elements of media content with the estimated profit that these elements of media content will create. Therefore, the operator should avoid that only very popular elements of media content are offered in order to satisfy a group of end users in order to control the costs of providing a set of elements to an end user. It is a fact that with a certain amount of very popular media content provided to end users of the media platform on demand, each additionally provided highly popular element of media content will generate less additional income than the additional costs associated with it. Therefore, the task of the media platform operator on demand is to carefully balance the set of elements. Accordingly, it is an object of the present invention to provide a method for compiling media content elements for an on-demand media platform that supports an on-demand media platform operator, maximizing its profit.

In addition, recommender systems have been used in various applications to recommend items (consumer goods, television programs, songs, etc.) or services to ease the problem of choice for end users who are faced with an overly large collection of items when choosing from them. The objective of such recommendation systems is to generate personalized recommendations on the elements that are supposed to please individual end users. To achieve this, information about the relevant end-users should be introduced into the recommendation systems. This information may include general information about the end user, for example, age, gender, hobbies and general interests, as well as information about his or her consumer behavior and take into account the opinion about certain elements and services. The collection of this information can be done indirectly by tracking and recording consumer habits of end users and generating purchase or rental history of an individual end user. In addition, end users should be asked to provide ratings specifically for recently consumed goods and services, so that the system can better study the preferences of individual end users. From this information, the recommender may receive items and services that individual end users may like. In addition, recommender systems can use statistics and personal information about the user to compare them with the data of other users and recommended elements that these users like or prefer.

An object of the invention is to provide a method for efficiently compiling media content elements for a media platform on demand, which can be performed automatically by a machine, such as a computer or the like. One aspect of the present invention is to use the information provided by the recommender systems about how each individual end-user is estimated to like each individual element of the proposed media content that is supposed to be added to the set of media elements content rented by a group of individual end users to determine which set of items should be offered in the next on-demand service period by a media operator. In addition, the present invention provides a combination of information generated by a recommender system with a rental predictor that predicts for individual end users the number of times that they will rent a media element of a potential set of media elements in the next maintenance period. Therefore, the method according to the present invention helps the media platform operator, upon request, to compile a balanced set of media content elements that will be offered for rent or sale to end users, the set of elements being selected to maximize profits as well as satisfy end users.

The above objective of the present invention is achieved automatically in accordance with the independent claim 1 of the claims and the system in accordance with the independent claim 15 of the claims. Within the scope of the invention, “automatic” means that the method can be performed by the system in whole or in part without the need for interaction with the operator. Further, it should be understood that all organizational steps regarding media content items relate to the metadata of the media content item, and not to the media content item itself.

According to the present invention, an automatic method is provided for automatically and efficiently compiling media content elements for a media platform on demand. The method comprises the steps of providing an on-demand media platform for renting media content items to a user or a plurality of on-demand media platform users and providing the media group of the media elements in question, including at least one media element in question . The considered elements of media content are considered to be offered to users on the media platform upon request for rental by users. The objective of the proposed method is the automatic determination and selection of suitable media content elements from the considered media group and the automatic provision of selected media content elements to the media platform upon request for rental to users.

The problem to be solved is a technical problem, consisting in the fact that a large number of elements of media content must be selected. For this purpose, a cost function is defined that takes into account the preferences of individual users. The technical aspect is to make more efficient use of storage space. For example, instead of providing mostly blockbusters and mostly popular videos, the storage space is optimized to provide a more balanced set of videos for rent.

At one stage, the recommender system automatically generates an estimated group of evaluated elements of media content by determining a user-specific rating of likes for each element of media content of the considered media group. A specific sympathy rating is generated relative to grading and / or rating based on the rental history of the given user and / or other typical media platform user upon request. When evaluating, for example, the media content elements that the user has rented can be classified according to characteristics, such as genre, actors, release date, content, etc., are provided with approximate ratings of likes and compared with similar considered media content elements. Matched about the considered elements of media content are automatically assigned the same rating of likes. Another possibility is that the user provides his personal rating of likes for elements of media content manually. In principle, the recommendation system according to the invention is configured to operate in accordance with the principles of operation of known recommendation systems.

At the next stage, the rent number predictor automatically estimates the user-specific number of media content elements, which the corresponding user of the media platform would rent from a nominal group of nominal media content elements for a specified period of time. Thus, the predictor of the number of rents takes into account information about the rental history of a given user and / or other typical users of the media platform upon request and the corresponding rating of likes of nominal elements of media content.

It is preferable that the considered media content elements that have a user-specific sympathy rating below a predetermined value are removed from the estimated group of evaluated media content elements of the recommendation system. By this, only media content items that a particular end user likes are counted for addition to the media platform upon request with respect to that particular end user only. Of course, relevant media content items may be considered for other end users of the media platform on demand, who like the particular media content item in question.

Preferably, the considered media content elements that the user has already rented are identified and removed from the group of considered media content elements by the recommender system. It is unlikely that the end user will rent the same element of media content twice within a relatively short period of time. Therefore, such considered elements of media content should not be considered to add to the set of media elements relative to the respective end user. Of course, relevant media content elements may be considered for other end users of the media platform upon request who have not yet rented the media element in question.

Advantageously, the account information about the user's rental history contains information about the genre, persons involved in creating a particular media content item, such as actors, artists, authors, musicians or directors, the release date of the media and / or the rental cost of the media content item. Thus, it can be analyzed which element of media content the individual end user prefers, and whether the individual end user has a preference for media content elements of a certain price class.

Preferably, the user-specific ranking of the considered media content items is generated by the recommender according to the corresponding user-specific rating of likes of the considered media content items. Using these tools, the considered media content elements are sorted by user-specific sympathy ratings for convenient identification of media content elements that an individual end user likes more (or less) than other media content elements.

Preferably, the considered media content elements with the worst user-specific sympathy rating are removed from the group of the considered media content elements by a recommender system.

Preferably, the specific media content items that the user is expected to rent from the group of media content items in question is determined by the number of rental predictor based on the user-specific number of media content items that the user will rent from the group of media items in question content and user-specific sympathy rating of media content elements. If the number of media content items that the end user is likely to rent is “n,” then elements from 1 to n media content with a better liking rating are likely to be rented by the respective end user.

An alternative forecasting model is also possible, where the similarity between the elements of media content is taken into account in the list of most likely rented videos for the end user. If the two elements of media content are very similar, it is less likely that the end user will rent both elements for a short period of time. To determine the likelihood that the end user will rent similar items for the same rental period, you can also take into account information from the rental history of the end users.

According to the invention, the rental cost of certain specific elements of media content from a respective content provider is preferably determined by a financial evaluation unit.

In addition, it is preferable that the turnover for renting certain specific elements of media content for the user is determined by the data processing unit.

It is preferable that the difference between certain costs and a certain turnover is calculated and the expected profit or loss is determined by the data processing unit. Based on this information, the method according to the invention can determine whether the set of the considered elements of the media content will generate a profit or loss. Comparing the expected profit or loss of various sets of considered elements of media content, the method may comprise, for example, automatically selecting the set of considered elements of media content that generates the largest profit.

Preferably, the method is performed for many different users. Certain information, such as a liking rating, the number of media items that users are likely to rent, and the number of specific media items that users are likely to rent, can be used to create averages for all end users in question. Thus, the operator is provided with information about the expected consumer behavior of the end-user group regarding a given set of elements from media content elements.

Preferably, certain specific elements of media content that many users are likely to rent are automatically added to the media platform upon request by a recommender system.

Mostly, the expected profit or loss is compared with a predetermined profit or loss. Thus, the operator of the media platform, on request, can understand whether the proposed set of elements fulfills these requirements or needs to be adjusted in order to achieve the goal. In addition, such regulation can be performed using a system that automatically performs the method in accordance with the invention. It is preferable that certain specific elements of media content, which are likely to be rented by many users, are added to the media platform upon request by the recommender system in case the expected profit is higher or equal to the set profit margin, or the expected loss is lower than or equal to the set loss amount.

Preferably, the media content items are digital media content items such as digital video, digital photos, digital music, computer programs, especially games and / or digital texts. In accordance with the invention, the media platform upon request aims to rent all types of digital media elements to end users, for example, by streaming from a server. Digital media elements may contain videos, games, music or books. Alternatively, media elements can be delivered to end users via email. In addition, the present invention also focuses on situations where end users buy items rather than rent items through an operator.

In addition, a system is provided for automatically executing a method for efficiently compiling media content elements for a media platform on demand according to the invention. The system comprises a recommendation block that is configured to determine a user-specific liking rating for media content items, a rental number predictor configured to determine the number of media content items that the user is expected to rent from the estimated group of media content items, and a block financial assessment, which is configured to generate a combined group of discrete elements of media content and to determine the expected profit or loss for ie the specific elements of media content. Each of the subsystems may contain a data processing unit. In addition, the system may include databases for storing and managing data related to the user and / or data related to media content. The database may, for example, contain metadata of media content elements that have been generated by a third party.

In one embodiment of the invention, the on-demand video platform operator defines a set of videos to potentially be added to the on-demand video platform. Information about the video set, such as, for example, names, genres, actors, publication time, nominations and / or awards, is transmitted to the rental income predictor. In addition, information about the actual and / or potential end users of the on-demand video platform end-user group, such as, for example, age, gender, and / or interests, is also provided to the rental income predictor. Information about end users can be stored in the database of end users and provided to the predictor of rental income directly from the database of end users.

The recommender system is provided with information about a set of videos and user ratings of likes of individual end users for videos from a set of videos. In addition, a recommender system may be provided with information on rental history of individual end users. Sympathy ratings can be provided by individual end users, explicitly or implicitly, by analyzing the rental history of end users. From this information, the recommender system generates sympathy ratings for each video from the video set and, therefore, ranking the video from the video set for each individual end user. In other words, the recommender system determines for each of the subscribers (end users) a list of videos from a given set of videos that he or she will like the most, possibly quantified with an appropriate rating of likes. This sympathy rating can, for example, be a value between zero and one, in which the number “one” means that the user rated this video as very liked, and “zero” means that the user does not like the video at all. Such a recommender system may be content based, or it may be based on collaborative filtering, or a combination of both.

Information about the ratings of likes for each video, the ranking of videos from a set of videos and the rental history of individual end users is provided to the predictor of the number of rents. In addition, the predictor of the number of rents is provided with information about the ratings of the likes of the video, which is considered in the history of rental of end users. The predictor of the number of rents determines for each of the end users an estimate of the number of videos that he or she will rent for the next service period. This estimate may simply be based on rental history, i.e. on information about how many videos the user has rented over the last periods of service, or for new users on the average number of rents (on average for all users or on average for all new users, based on information from past service periods), not taking into account which video set is offered for the next service period.

Alternatively, the rental number predictor can use the ranking of videos from a given set of videos for an individual end user, as indicated by the recommender system, to estimate the number of rents or, if required, to evaluate the exact set of videos that an individual end user is expected to rent.

In addition to using it to predict the number of rentals, the rental history can also be used to understand which videos the end user has rented in past maintenance periods. Videos that were recently rented will probably not be rented again in the near future.

The assessment of which video this user is likely to rent in the upcoming maintenance period can now simply be determined as follows. A list containing the video rankings for the individual end user is retrieved from the recommender system, and videos that were recently rented are removed from this list. Then, assuming that the estimated number of rents for the upcoming service period is k, the estimate of the set of videos that this user will rent in the upcoming service period is simply determined by the first k elements from the received adapted list.

In a more advanced embodiment, the estimated number of rented videos may be affected by the estimated liking rating indicated by the recommender system. In this embodiment, the recommender system determines a first list containing the rankings and likes ratings of the videos that an individual end user has rented during a certain service period in the past. Ratings of likes of videos provided during the same service period that were not rented can also be viewed in the first list.

In addition, a second list of estimated ratings and ratings of likes of the video from the proposed set of videos to offer end users in the next service period is provided. The sympathy ratings of the various rankings of the first and second list are compared with each other. If the liking ratings of the videos from the second list are higher or equal to the videos from the first list that have been rented by the end user in the past, the video is likely to be rented by the end user with a higher probability if the liking rating is the same, or lower than the liking rating of a video that the end user has not rented in the past. This information is statistically evaluated to generate an estimate of the number of videos that the user will rent from the intended set of videos.

If the system uses a differential pay-per-view pay-per-view model, the user’s rental history can be further analyzed to see if the user mainly rents more expensive, recently released videos, or if the user rents mostly less expensive videos. This can be used to have a more accurate forecast of the amount of money that the user is expected to spend on video rental in the upcoming service period. In addition to the number of videos that the end user is likely to rent from the set of videos offered in the next maintenance period, the number of rent predictor can also evaluate which videos the end user is likely to rent. This rating is based on the thesis that the end user will rent a video according to the ranking regarding the rating of likes. The end user is likely to rent videos with the highest liking ratings. If the number of rated videos that the end user was planning to rent was “k,” the end user will most likely rent the first k videos from his individual ranking.

An important feature of the present invention is a rental income predictor that estimates the income that a given set of media will generate when offered to a group of end users over the next service period. The rental income predictor uses a recommender system and the rental predictor. Therefore, the recommendation system and the predictor of the number of rents can be considered as subsystems of the predictor of rental income. Provided that the recommender system gives for each combination of the end user and set of videos the ranking of the videos, and provided that the predictor of the number of rents gives for each combination of the end user and set of videos an estimate of the number of films that this user is most likely to rent in the next period maintenance, it is easy to calculate the expected amount of money that this user will spend on rent in the next maintenance period. Perhaps the rental history of this user is taken into account.

Now, the full rental income can simply be obtained by summing up for all users the expected amount of money that each user will spend on rent in the coming period. Obviously, as an alternative, more advanced probabilistic models can also be used to predict total rental income. A more advanced embodiment of the rental income predictor also includes taking into account the popularity of the individual videos. For example, videos that were recently released as feature films and attracted many viewers in movie theaters and / or were well received in the views are likely to attract more interest from end users. A possible way to include these popularity indicators is to further increase the liking rating of these popular elements, compared to the estimated degree of sympathy given by the recommendation system.

It should be noted that determining the popularity of a new video is not the same as determining a user-specific liking rating, and how often a particular video will be rented is not the same as evaluating the user-specific number of media content items that the user will to rent.

In addition, if the resulting degree of sympathy for these popular elements is large enough for a given end user, then the estimate of how much rent the end user is likely to have in the upcoming service period will increase accordingly. The exact value of this fraction can be determined experimentally. Accordingly, ratings of video likes that, for example, received negative reviews from media critics, can be negatively impacted. It can also negatively affect the number of videos that it is estimated that the end user will rent and should be accounted for accordingly.

The rental income predictor can be used as a simple tool, where the operator can repeatedly set the video set and the subscriber set as input to receive the expected video rental income for these given inputs as output. In addition, the predictor of rental income can be part of a large system, where the large system automatically tries to optimize net profit also taking into account the rental costs that the operator must pay for this set of videos. In this case, various combinations of video sets can be generated, and an estimated net profit can be calculated for each of these combinations. This large system can simply return the highest n combinations that are expected to generate the largest net rental income. All combinations could be considered in an exhaustive way, but this can be impractical.

Alternatively, a local search approach can be applied to repeatedly try to find the best solution, trying to add or skip a set of videos for the current combination, whenever it improves the overall result. Then, in the end, the operator can choose one of these combinations, taking into account, possibly, other aspects, such as strategic considerations, as to whether the operator wants to invest in expanding its subscriber base.

Finally, the above system can also be used to measure customer satisfaction. From the estimated liking rating of videos that are estimated to be rented by a given end user, an estimate of how satisfied the end user is can be obtained. In addition, if the estimated number of rents for a user increases or decreases compared to the (average) number of rents during past service periods, this can be interpreted as a sign that this user's satisfaction has increased or decreased, respectively.

A preferred embodiment of the invention relates to a system and / or method for efficiently compiling a data set, in particular a set or group of several media content items and / or data, in particular information about media content items. Elements of media content are, for example, digital and / or non-digital elements, for example, DVD, CD, books, movie files, music files and / or electronic book files and so on. Data is stored in a database, such as a server or computer. Data is part of the on-demand media platform for leasing media content to a user, for example, via a website, by phone, or by a suitable TV channel. In addition, the data contains information about the rental history of each user and / or individual user ratings for media content items. The user's rental history contains additional information about specific elements of media content, such as, for example, name, genre, actors, rental price, release date or the like. Information is stored as metadata in the user database, which is the preferred part of the database. The system and method are configured to provide an additional set or group of data under consideration, wherein the additional data set contains at least one media content element, and the considered media content elements are considered in order to be proposed, namely, presented for a particular Users using the media platform upon request. To ensure, in particular, the generation of an additional data set, the metadata of the media content elements stored in the database are provided, which means they are transferred to the recommendation system, and the recommendation system receives metadata or information about user likes, dislikes and / or consumer behavior stored in the form of metadata in the user’s database, while the high sympathy rating of an element means that the user likes the specific element of media content more than the element of media A low-rated online content. As a result, and based on the receipt of metadata from media content elements, the recommender system generates an additional data set, in particular, estimated groups of evaluated media content elements by determining a user-specific sympathy rating for each considered media content element, and a certain sympathy rating is generated relative to the certification and / or estimates based on the rental history of a particular user and / or other characteristic user of the media platform upon request. In addition, the system and method is configured to estimate or predict a user-specific number of media content elements, which a particular user of the media platform will, upon request, rent from an estimated group of evaluated media content elements for a certain period of time, as a predictor of the number of rents, taking into account historical information rental of a specific user and / or other characteristic users of the media platform upon request and the corresponding rating of likes of the estimated elements m dia content. Thereafter, an estimated, which means predicted, user-specific quantity is provided to the media platform operator on demand in order to regulate the storage of multiple media content items on the media platform on demand. The above system and method provides an improved system to provide more efficient use of digital (e.g., memory) and / or non-digital (e.g., shelf) storage space because an operator using the above system and method is able to estimate how much storage space is needed for a certain period of time in the future. For example, instead of simply offering mostly blockbusters and mostly popular videos, the storage space is optimized to provide a more balanced set of videos for rent.

The invention will now be described by way of example with reference to the drawing. In the drawings:

FIG. 1 describes a selection system for efficiently compiling media content elements for an on-demand media platform.

Referring to FIG. 1, an on-demand media platform 10a is provided comprising a plurality of proposed media content elements 12 for a user 40. Media content elements 12 may include digital and / or non-digital media content elements. User 40 may rent the proposed media content elements 12 in accordance with the rental model as described above, for example, pay per view or monthly subscription with a limited or unlimited number of rental media content elements 12. The rental history of user 40 and individual user ratings for media content elements 12 are stored in the user database 42. The rental history of user 40 may contain information about specific media content elements 12, such as name, genre, actors, rental price, release date, or things like that. In addition, the number of each media content item 12 rented by the user 40 and the user's personal data can be stored in the user database 42.

The content provider 30 offers the group 10b of the considered media content elements 12 to the operator of the media platform 10a upon request. Alternatively, a plurality of content providers 30 can be considered. The considered media content elements 12 are further provided to the selector system 14 for efficiently compiling media content elements 12 for the media platform 10a upon request. Selector system 14 comprises a recommender system 16, a predictor 18 of the number of rents, and a financial valuation unit 20. Each of these components contains a data processing unit not shown.

The metadata of the considered media content elements 12 is provided by the recommender system 16. The recommender system 16 obtains information about likes, dislikes, and consumer behavior of user 40 from user database 42. In addition, many other sources can be examined to determine the likes and dislikes of user 40, for example , sales metrics, market research, or the like. Then, the recommender system 16 evaluates the information obtained to determine the sympathy rating for each considered element 12 of the media content. In this regard, a high sympathy rating of an element means that the user 40 likes the particular media content element 12 more than the media content element 12 with a lower sympathy rating. The range of ratings can, for example, range from the fact that an element is really like to the fact that the element is not at all like it. As a result, an estimated group 10c of user-rated media content elements 12 is generated for each individual user 40 individually. Alternatively, only a given user group 40 can be considered in this regard.

The rating of likes may also be manipulated by the operator, for example, if the needs of a specific target group of a potentially new user should be satisfied by the media platform 10a upon request. In addition, the recommender system 16 may sort the estimated media elements 12 of the evaluated group 10 with a sympathy rating value. The liking rating of the rated media content elements 12 that have already been rented by the user 40 can be automatically lowered by the recommender system 16, since it is unlikely that the user 40 will again rent the same media content item 12. The time interval between the last rental and the present time can also be considered, and the decrease in the rating of likes will be less than the longer the corresponding period of time. As an alternative to lowering the sympathy rating, the corresponding rated media content element 12 can be removed from the rated group 10c by the recommender system 16. In addition, the rated media content elements 12 that have a sympathy rating below a predetermined threshold, for example, elements that the user 40 do not like, can also be removed from the rated group 10c by the recommender system 16. The predictor 18 of the rental number compares information from the user database 42 with the estimated elements 12 of the media content from the rating constant group 10c for determining the number of items of media content that the user 40 is likely leases from the estimated group 10c valued elements 12 media content for a certain period of time, e.g., weeks or months. Thus, the rental history of user 40 is especially considered. In the event that user 40 has rented a certain number of media content elements 12 for a certain period of time in the past, it is likely that the same user 40 will rent approximately the same number of media elements 12 content 12 for the same period in the future when the sympathy rating of newly provided media content elements 12 is approximately the same as the sympathy rating of already rented media content elements 12.

If the sympathy ratings of the newly provided media content elements 12 are higher than the sympathy ratings of the already rented media content elements 12, the expected rental amount may also be higher. Accordingly, if the sympathy ratings of the newly provided media content elements 12 are lower than the sympathy ratings of the already rented media content elements, the expected rental amount may be less. For this consideration, the average value of the sympathy ratings and / or the sympathy ratings of the least liked media content elements 12 rented by the user 40 can be considered. As a result, the rent predictor 18 provides a quantitatively defined group 10d of quantitatively defined media content elements 12 separately for each individual user 40. Alternatively, to create a quantitatively defined group of 10d quantitatively defined elements of media content, the predictor is 18 ch SLA rental can only consider a given group of characteristic users 40.

In a subsequent step, the financial evaluation unit 20 combines all the quantitatively defined groups 10d for each user 40 in question into a combined group 10e of combined media content elements 12. The financial evaluation unit 20 compares rental costs from the content provider 30 for each combined media content item 12 of the combined group 10e with the expected rental income of the combined media content item 12 to users 40. The combined media content items 12 that incur higher costs than income, are automatically removed from the combined group 10e by the financial evaluation unit 20. Thus, the financial evaluation unit 20 generates a selected group 10e of selected media content elements 12 and provides the selected group 10e to the media platform 10a upon request. This can be done either by simply adding the selected media content elements 12 to the media platform 10a upon request, or by replacing the proposed media elements 12 upon request of the media platform 10a upon request with the selected media content elements 12.

In addition, a selected group of selected media content items may be offered to the operator for human control. Different groups of samples can then be evaluated by the operator.

The following describes the implementation details of the financial evaluation unit in accordance with the invention.

Revenue modeling for a given selection of video sets

In this example, the media content items are video. However, instead of video, a different kind of media content element can be considered in this way. Consider N content providers numbered 1, 2 ... N, each of which offers sets of videos that a video on demand (VoD) provider can selectively rent for a period of time. For this choice, S _{11} , S _{12} , ..., S _{{1, n (1)}} , S _{21} , S _{22} , ..., S _{{2, n (2)}} , S _{{N , 1}} , S _{{N, 2}} , ..., S _{{N, n (N)}} , then we express the income that the VoD provider, having a subscriber base designated U, expects to generate. Let O denote the full sentence of the video, i.e.

For each user u E U and video v E O, the recommender will provide a degree of sympathy l (u, v) E [0, 1], indicating how much user u likes video v. Based on the proposal, the probability P (0, u, v) that user u rents video v in the coming period can be estimated as follows. Suppose a user rents the number of n (u, O) videos for a period, which may depend on the (size) of the O offer. This can be easily estimated using the user's u rental history. Then, for example, a random sample of size n (u, O) is selected from a subset of size M from the video to O with the highest degree of sympathy. Alternatively, suppose a user rents a number of n (u, O, g) videos for a period of a particular genre g. This can be easily estimated using the user's rental history. Then, for example, a random sample of size n (u, 0, g) is selected from a subset of size M of the video at 0 with the highest degree of sympathy and having the genre g. In general, if a user is characterized by many personal channels (Pronk, V., J. Korst, M. Barbieri & A. Proidl [2009]. Personal television channels: simply zapping through your PVR content, Proceedings of the 1st International Workshop on Recommendation- based Industrial Applications, in conjunction with the 3rd ACM Conference on Recommender Systems, RecSys 2009, New York City, NY), then the definition filter for each channel can be used instead of specific genres to subdivide the video selection. In addition, each channel should be equipped with a separate recommender, which provides an appropriate degree of sympathy. If a video should be selected in the context of two or more channels, then only the one with the highest degree of sympathy is retained, and alternatives are selected for other channels. In addition to each of the indicated methods of constructing a set of videos, this design can be refined by replacing selected videos that appear in the blacklist, for example, from a recently rented video, with appropriate alternatives.

In any of the examples described above, a set (i.e., a selected group) of size n (u, O) videos is selected. These selected videos get a high probability of P (O, u, v), for example, probability 1, and any other videos in O - a low probability of P (O, u, v), for example, probability 0.

If the price u must pay for rent v is p (u, v), and the subscription fee for this user for the rental period is determined by s (u), then the total income that the VoD provider receives from u for the rental period, provided 0 can be estimated by

Summarizing the contributions from all users u to their customer base U and all sets, the total income for the VoD provider over the coming period, including the periodic subscription fee s (u), can be expressed as

To promote diversity, in order to preserve the customer base, the deregularization factor can be included in the above expression. In particular, this expression can be multiplied by

where R is a monotonically non-decreasing function from natural numbers to the real interval [0,1], and | O | denotes the number of elements O. This function works as a fine on income, and can be studied by historical analysis of the size of the client base, depending on the total size of the video offer. As a simple example, the function R only reaches the values 0 and 1, which means that the general sentence must have some minimum size.

On the other hand, the costs associated with renting and storing the sets S _{i1} , S _{i2} , ..., S _{{i, n (i)}} from the content provider i can be quantified as

C _i ({S _{{ i , 1}} , S _{{ i , 2}} , ..., S _{{ i , n ( i )}} }).

The summation of costs for all content providers then forms the full costs.

To summarize, the revenue of the VoD provider for the next period is set by

It goes without saying that costs that are not dependent on video sets should not be included in the equation.

The ability to refine the calculation of total revenue is to introduce a price-dependent selection algorithm. Then, assuming that the total pricing is denoted by p, the number n (p, u, O) of the video is selected, and each of the described selection processes becomes a little more complicated, since, for example, the maximum amount spent over a period can then be taken into account .

In the automatic optimization procedure, the expected income from the various options for choosing video sets is calculated to determine the set of videos that promises the highest income in order to change the choice of video sets, a local search for the options for choosing video sets can be automatically performed as follows. Now, the local search approach is as follows. For each choice, S _{11} , S _{12} , ..., S _{{1, n (1)}} , S _{21} , S _{22} , ..., S _{{2n (2)}} , S _{{N, 1 }} , S _{{N, 2}} , ..., S _{{N, n (N)}} , a neighborhood is defined. This neighborhood defines choices that are very similar but different from this choice, for example, because one set has been added or deleted. Using one or the other, respectively, determining the neighborhood function, a local search can begin in an arbitrarily selected initial choice, and the revenues are calculated for this choice. This initial choice is the current choice. Then all or some of the neighbors of the current selection are pre-selected and their incomes are calculated. If among these choices there are choices that result in more revenue than those made by the current selection, one of these choices is defined as the current selection and improvement is achieved. This process of generating improvements is iteratively repeated until a local minimum is reached and no improvement can be made this way anymore.

This iterative improvement process can be repeated using alternative initial selections and the best local minimum will be selected.

It is also possible to repeat the local search by selecting different pricing to maximize profits. Pricing for research can be set in advance taking into account competition.

The specific local search described above is for reference only. It will be apparent to those skilled in the art that there are numerous variations of local search that are well described in the literature, such as forbidden search, stochastic local search, genetic local search, simulated annealing, etc., and that optimization problems can be considered numerous variations.

To ensure that the VoD provider will always have the choice of the most popular video available for rent, they could be excluded from the consideration above, for example, by removing them from the video sets. The optimization solution described above will then concentrate on the video, for which individual differences in degrees of sympathy are seen more prominently.

Claims (34)

1. A method of selecting a data set in the form of a group of digital media content elements from a plurality of groups of digital media content elements for an offer on a media platform on demand, in order to adjust the memory for elements of media content on a media platform on demand, comprising the steps of: - providing a media platform on demand to offer groups of media content elements for rent by users of a media platform on demand, and the media platform on demand allows you to broadcast digital elements of media content from a media server on demand to users through a data network, - providing the metadata recommendation system of the considered elements of media content forming a plurality of considered media groups, each considering the media group of the considered elements of media content contains a plurality of considered elements of media content, and the considered media groups of the considered elements of media content are considered to be offered on a media platform upon request, - generation by the recommender system of evaluated groups of estimated media content elements by determining, for many different users, a user-specific sympathy rating for each media element under consideration, and a user-specific sympathy rating is determined taking into account certification or assessment based on the rental history of a particular media platform users upon request, while the rental history and user-specific likes rating are stored in the database, The rental history contains information about the elements of media content in the form of metadata, - estimating by the predictor the number of rents, which is implemented in the data processing unit, for each of the evaluated media groups and for each of the many different users, the corresponding user-specific number of estimated elements of media content that the corresponding user of the media platform would rent from the corresponding an estimated group of estimated elements of media content over a certain period of time, using information on the rental history of a particular platform user m dia on request and the corresponding rating Award valued elements of media content, - using the sum of user-specific quantities of estimated media content elements that the corresponding user of the media platform would rent upon request from the corresponding evaluated group of evaluated media content elements for a certain period of time, and using the specified pricing to estimate the estimated revenue for each of the estimated groups rated media content elements, and - selection from the evaluated groups of the estimated media content elements of that evaluated group of the estimated media content elements that is associated with the largest revenue among the estimated revenue for an offer on the media platform upon request. 2. The method according to p. 1, in which: - the aforementioned stage of providing the metadata recommendation system of the considered elements of media content contains an additional step of compiling the considered media groups, - the recommendation system removes those elements of media content from the considered media groups that have a user-specific sympathy rating below a predetermined value. 3. The method according to p. 1, containing an additional step: - identification and removal of the considered elements of media content that a particular user has already rented from the considered media group of the considered elements of media content through a recommender system. 4. The method according to p. 1, in which the information on the rental history of a particular user contains information about the genre, the persons involved in creating the media content item, the release date of the media and / or the rental cost of the media content item. 5. The method according to claim 1, containing an additional step: - generating a user-specific ranking of the estimated elements of the media content through a recommender system in accordance with the corresponding user-specific rating of likes of the evaluated elements of the media content. 6. The method according to p. 5, containing an additional step: - removal of rated content items with the worst user-specific rating of likes from the group of rated media content items through a recommender system. 7. The method according to p. 1, containing an additional step: - generating a quantitatively defined group of quantitatively determined media content elements by determining specific estimated media content elements that a particular user is expected to rent from the estimated group of estimated media content elements, using a rental number predictor based on a specific user-specific number of media elements content that the user will rent, from the rated group and specific user-specific ratings with mpaty priced items of media content. 8. The method according to p. 7, containing an additional step: - determining the costs for renting each quantitatively determined element of media content from a quantitatively defined group from the corresponding content provider through a financial assessment unit. 9. The method according to p. 8, containing an additional step: - determining the turnover for renting certain specific elements of media content for a particular user through a financial assessment unit. 10. The method according to p. 9, containing an additional step: - calculating the difference between certain costs and a certain turnover and determining the expected profit or loss through the financial assessment unit. 11. The method according to claim 1, in which the elements of the media content of a quantitatively defined group of each user in question are combined into a combined group of elements of media content by means of a financial evaluation unit. 12. The method according to claim 10, in which the elements of media content of a quantitatively defined group of each user under consideration are combined into a combined group of elements of media content by means of a financial evaluation unit, and the expected profit or expected loss is compared with a predetermined profit value or loss value by means of a financial unit assessment. 13. The method according to p. 12, containing the stage: - adding the combined specific elements of media content that many users are likely to rent to the media platform upon request through a recommender system, if the expected profit is higher or equal to the set profit value or the expected loss is less than or equal to the set loss value. 14. The method of claim 1, wherein the media content items are digital media content items, digital video, digital photos, digital music, computer programs, or digital texts. 15. A system for automatically executing a method for selecting media content elements for an offer on a media platform by request according to claim 1, comprising: - recommender block for determining user-specific sympathy ratings for media content elements; - a predictor of the number of rents to determine the number of media content elements of a user who is expected to rent from an estimated group of media content elements; and - a financial assessment unit for generating a combined group of quantitatively defined elements of media content and determining the expected profit or loss for the presentation of specific elements of the media context.

Images